TY - JOUR T1 - Ascertainment of cause-specific mortality in COPD: operations of the TORCH Clinical Endpoint Committee JF - Thorax JO - Thorax SP - 411 LP - 415 DO - 10.1136/thx.2006.072348 VL - 62 IS - 5 AU - Lorcan P McGarvey AU - Matthias John AU - Julie A Anderson AU - Michael Zvarich AU - Robert A Wise Y1 - 2007/05/01 UR - http://thorax.bmj.com/content/62/5/411.abstract N2 - Background: TORCH (Towards a Revolution in COPD Health) is an international multicentre, randomised, placebo-controlled clinical trial of inhaled fluticasone propionate/salmeterol combination treatment and its monotherapy components for maintenance treatment of moderately to severely impaired patients with chronic obstructive pulmonary disease (COPD). The primary outcome is all-cause mortality. Cause-specific mortality and deaths related to COPD are additional outcome measures, but systematic methods for ascertainment of these outcomes have not previously been described. Methods: A Clinical Endpoint Committee (CEC) was tasked with categorising the cause of death and the relationship of deaths to COPD in a systematic, unbiased and independent manner. The key elements of the operation of the committee were the use of predefined principles of operation and definitions of cause of death and COPD-relatedness; the independent review of cases by all members with development of a consensus opinion; and a substantial infrastructure to collect medical information. Results: 911 deaths were reviewed and consensus was reached in all. Cause-specific mortality was: cardiovascular 27%, respiratory 35%, cancer 21%, other 10% and unknown 8%. 40% of deaths were definitely or probably related to COPD. Adjudications were identical in 83% of blindly re-adjudicated cases (κ = 0.80). COPD-relatedness was reproduced 84% of the time (κ = 0.73). The CEC adjudication was equivalent to the primary cause of death recorded by the site investigator in 52% of cases. Conclusion: A CEC can provide standardised, reliable and informative adjudication of COPD mortality that provides information which frequently differs from data collected from assessment by site investigators. ER -